The quality of data contained in accounting information systems has a significant impact on both internal business decision making and external regulatory compliance. Although a considerable body of literature exists on the issue of data quality, there has been little research done at the task level of a business process to develop effective control strategies to mitigate data quality risks. In this paper, we present a methodology for managing the risks associated with the quality of data in accounting information systems. This methodology first models the error evolution process in transactional data flow as a dynamical process; it then finds optimal control policies at the task level to mitigate the data quality-related risks using a Markov decision process model with risk constraints. The proposed Markov decision methodology facilitates the modeling of multiple dimensions of error dependence, captures the correlated impact among control procedures, and identifies an optimal control policy. A revenue realization process of an international production company is used to illustrate this methodology.